Demand Has Gone Parabolic: $82B Revenue, $91B Q2 Guide, Vera CPU Opens a $200B TAM, Anthropic Joins the Frontier Roster — Muted Reaction Is Positioning, Not Thesis

Results vs. Consensus

YoY Comparison

QoQ Comparison

Revenue Assessment

$81.6B revenue in a single quarter places NVDA at a ~$330B annualized run-rate revenue base, on track for FY27 to clear $360-380B. The +20% sequential growth on a $68B Q4 FY26 base is the most striking element — at this scale, +20% QoQ is mathematically extraordinary; it implies ~$13.5B of net new quarterly revenue created in 13 weeks, more than ADI's annual revenue. The 14-quarter sequential growth streak is now itself a structural data point: the demand surface for accelerated compute is so broad and so unsaturated that NVDA has not had a down sequential quarter in three-and-a-half years through multiple architecture transitions (Hopper → Blackwell → Blackwell Ultra → Rubin coming). The Q2 guide $91B implies another +$9B sequential (+11% QoQ) — still above-seasonal at this scale and above buy-side whisper $90B. Annual revenue trajectory now tracks $360-400B FY27, $500-600B FY28 if Vera Rubin ramps as guided and Vera CPU $200B TAM begins to monetize.

Margins Assessment

Non-GAAP gross margin 75.0% sequentially flat against Blackwell at scale is the operational achievement that the headline numbers can obscure. The Blackwell architecture transition was modeled to compress gross margin into the low-70s during the ramp; instead, the steady-state has settled at 75%, and management guided 74.9% Q2 with the full year "in the mid-70s." The implication: Blackwell incremental gross margin is actually higher than Hopper steady-state because (a) the platform sells systems (not just chips), (b) the networking attach rate ($15B Q1, +200% YoY) is structurally higher-margin, and (c) the AI factory ROI per dollar of capital is rising as model efficiency improves and rental rates rise (H100 +20% YTD, A100 +15% YTD). Operating margin tracks ~65% non-GAAP; the only constraint is OpEx growth, which management guided to "upper 40s" YoY — meaningful R&D investment but well-contained relative to the revenue trajectory.

EPS Assessment

Non-GAAP EPS $1.87 is +95% YoY off the Q1 FY26 base of $0.96; reported beat Street $1.76 by +6.25%. The tax rate at 16% (vs guide of 17-18%) added roughly $0.04-0.05 to the beat — so the operational beat is $0.06-0.07 / +4% above Street. FY27 EPS trajectory now tracks $7.50-8.00 (vs prior Street ~$6.50) on the combination of (a) Q1 + Q2 guide running ahead, (b) Vera CPU $20B FY27 contribution layering on, (c) Anthropic capacity expansion adding incremental Q3-Q4 demand, (d) lower tax rate guide (16-18% vs prior 17-19%), and (e) buyback acceleration reducing share count by ~2-3% over FY27. At $7.75 mid-range FY27 EPS, current $224 stock prints 29x forward P/E — historically the mid-quartile for NVDA, not the bottom-quartile. The multiple-expansion case requires continued FY28 revenue growth visibility, which Vera Rubin and Vera CPU + sovereign acceleration are providing.

Segment Performance — New Disclosure Framework

Management introduced a new segmentation framework with this print: two market platforms (Data Center + Edge Computing); within Data Center, two submarkets (Hyperscale + ACIE). Nine quarters of revenue restated under the new framework are posted on the IR site. This is the most material disclosure change since the 2017 segmentation shift and reframes how the Street will model NVDA going forward.

Data Center — Hyperscale

$38B (~50% of Data Center, +12% QoQ). Hyperscale captures the "big six" public-cloud + largest consumer-internet customers. The +12% sequential is meaningful — at $38B base, it implies ~$4B of net new quarterly hyperscale revenue. The composition: hyperscalers are deploying for three concurrent use cases (per Jensen): (1) data processing + ML workload acceleration migrating CPU-to-GPU, (2) internal AI services for their own products, (3) public cloud capacity to rent to NVDA-ecosystem customers. The structural framing — hyperscaler CapEx track $1T annual by 2027 — supports continued +10-15% QoQ growth through FY27 in this segment alone.

"Microsoft's Farweave the world's most powerful AI data center is now live, ahead of schedule, powered by hundreds of thousands of Blackwell GPUs. Starting this year, AWS will add more than 1 million Blackwell and Rubin GPUs and are collaborating on spectrum networking. At Google, Blackwell will be offered to customers in the cloud including confidential computing capability."
— Colette Kress, CFO

Assessment: Hyperscale at $38B / +12% QoQ is the steady-compounder leg of the story now. Growth here is throttled by power, real estate, and supply (NVDA's own constraints), not by demand. The Microsoft Farweave / AWS 1M+ GPU / Google confidential-computing announcements are the proof points that the big-six are at full deployment cadence. This is the bedrock of the FY27 revenue base.

Data Center — ACIE (AI Clouds / Industrial / Enterprise)

$37B / +31% QoQ. ACIE — the new submarket — captures AI cloud providers (CoreWeave, Lambda, Crusoe, etc.), sovereign AI clouds, on-premises enterprise, and industrial AI infrastructure. AI cloud revenue alone "more than tripled YoY" per Colette. The +31% QoQ growth is materially faster than hyperscale's +12% and is the proof point of Jensen's "second category" framing he has been describing for several quarters. Number of partner data centers exceeding 10MW has nearly doubled in one year to over 80 sites.

"ACIE revenue was $37 billion and grew 31% quarter over quarter including AI cloud revenue that more than tripled year over year. Our customers have enabled rapid stand up of AI compute capacity. The number of partner data centers exceeding 10MW has nearly doubled in just 1 year. Now surpassing 80 sites. Sovereign revenue increased more than 80% year over year. NVIDIA AI infrastructure is now deployed across nearly 40 countries representing $50 trillion in GDP."
— Colette Kress, CFO

Assessment: ACIE is the under-modeled growth engine the Street has been struggling to size. At $37B Q1 / +31% QoQ, the segment annualizes to ~$170B if growth rates moderate to 20% QoQ exit-rate — versus virtually nothing 2 years ago. This is structurally the most differentiated revenue pool because (a) hyperscaler share gains are capped by big-six relationships, (b) ACIE customers have less negotiating leverage and need NVDA's fully integrated platform solution, (c) sovereign AI is a multi-year multi-country pipeline that's still in early-quarter ramp. ACIE could be larger than Hyperscale by FY28 if current trajectories hold.

Data Center — Compute vs. Networking

Data Center Compute $60B (+77% YoY); Data Center Networking $15B (+200% YoY, nearly tripled). Networking is now ~20% of Data Center revenue and growing materially faster than compute — InfiniBand grew 4x YoY on next-generation XDR deployments; Spectrum-X Ethernet platform "now larger than all Ethernet network peers combined." The networking attach rate is the gross-margin-supportive structural mix shift; networking is higher-margin than compute and the +200% YoY trajectory protects consolidated gross margin even as compute volumes ramp.

Assessment: Networking re-rating is the most under-modeled gross-margin lever. At $15B Q1 / +200% YoY, networking annualizes to ~$60-70B FY27 revenue — comparable in scale to the consolidated revenue of most large semiconductor companies. The Spectrum-X positioning ("larger than all Ethernet network peers combined") makes NVDA the dominant AI-networking infrastructure provider, displacing legacy Cisco/Arista at the AI factory layer. Gross margin tailwind is structural.

Edge Computing

$6.4B (+10% QoQ / +29% YoY). New platform encompassing PCs, gaming consoles, workstations, AI RAN base stations, robotics, automotive. Blackwell workstation demand was the strong contributor; consumer (gaming GPU) demand fell modestly on higher memory + system prices. Physical AI revenue (subset of Edge) exceeded $9B over the trailing 12 months.

"Our partnership with Uber will power the robotaxi fleet across nearly 30 cities and 4 continents by 2028. And in robotics, leading companies across a range of industrial, surgical, and humanoid applications are building on NVIDIA's technology, to develop and deploy at scale."
— Colette Kress, CFO

Assessment: Edge is the smallest platform but the highest optionality leg. Physical AI / robotics is still pre-inflection (NVDA gets the design wins now; revenue lags 18-36 months). Uber robotaxi (30 cities / 4 continents by 2028) is the single most quantifiable physical AI deployment timeline; humanoid robotics + industrial robotics are multi-year revenue layers. Edge is not material to FY27 modeling but could be 15-20% of revenue by FY30 if physical AI inflects as Jensen describes.

Key Topics & Management Commentary

Overall Management Tone: The most assured posture we have seen from Jensen on any NVDA earnings call. Every framing was forward-leaning — "demand has gone parabolic," "AI is no longer a nice to have, it is a necessity," "the world is rebuilding computing for agentic AI" — without the hedging or caveat-language that typically accompanies management commentary at extreme scale. Notably, Colette's prepared remarks were tightly disciplined (record print, record FCF, record returns), while Jensen's Q&A answers were strategically expansive (new segmentation framing, Vera CPU TAM, agentic AI inflection thesis). The combination — disciplined financials + bold strategic framing — is the posture of a company that believes it is in an unprecedented secular position and is building the investor narrative to match.

1. The Vera CPU Launch — A $200B TAM New to NVIDIA

The most material new disclosure of the call. NVDA unveiled the Vera CPU, a custom-Arm-core CPU codesigned end-to-end with Rubin GPUs and NVLink. Specifications: 1.5x performance per core, 2x performance per watt, 4x density per rack vs. x86 alternatives. Designed specifically for agentic AI workloads where token-throughput-per-dollar is the operative economic metric (vs. cores-per-dollar in the cloud-rental era). $20B FY27 visibility for standalone CPU revenue; $200B TAM for the CPU market NVDA has never previously addressed.

"Vera opens a brand new $200 billion TAM for NVIDIA, a market we have never addressed before. And every major hyperscale and system maker is partnering with us to get it deployed. We have visibility to nearly $20 billion in total CPU revenue this year. Setting us up to become the world leading CPU supplier."
— Colette Kress, CFO

Assessment: Vera CPU is the most significant strategic disclosure since the original H100 launch. The market opportunity is structurally different from the GPU franchise: agentic AI workloads need CPU horsepower for orchestration, tool use, memory management, and sub-agent coordination — Jensen described the future state as "billions of agents using PCs" running on CPUs. At $20B FY27 standalone revenue and $200B TAM, Vera CPU adds a third growth vector (after Data Center GPUs and networking) that the Street has not yet modeled. Importantly, Vera CPU is NOT included in the $1T Blackwell+Rubin CY25-27 visibility framework — it's incremental. The market-share dynamic vs. Intel/AMD is the open question, but with hyperscalers + system makers already partnering on deployment, NVDA is positioned to take meaningful share at the agentic-AI CPU socket.

2. Anthropic Strategic Partnership — Frontier-Model Coverage Completes

Anthropic was disclosed as a new strategic partner for capacity expansion across AWS, Azure, CoreWeave, and xAI. Previously NVDA's coverage of Anthropic was "largely zero" — Anthropic had been running primarily on AWS Trainium and Inferentia. The shift to NVDA capacity expansion for Anthropic is meaningful because it completes the frontier-model partner roster: Anthropic + OpenAI + Gemini + xAI + Meta + Microsoft AI + MSL + TML + Reflection + Perplexity + Cursor.

"We have deepened our collaboration with Anthropic and are delighted to be a strategic partner to expand their compute capacity. We will support the company's growth trajectory through AWS, Azure, CoreWeave, StacyX AI, and more. Now with the addition of Anthropic, to OpenAI, Gemini, StacyX xAI, Meta, MSL, Microsoft AI, TML, Reflection, Perplexity, Cursor, and other major frontier labs already building on NVIDIA. Our share of frontier AI models will grow significantly."
— Colette Kress, CFO

Assessment: The Anthropic partnership is materially positive on two dimensions: (1) Anthropic's Claude model is now growing at "incredible pace" with "breakout growth in OpenAI's codex since launch of GPT 5.5" — capturing that growth on NVDA infrastructure rather than ASIC alternatives validates the platform thesis, (2) Anthropic's previous-quarter share-loss to Trainium reverses, removing the most visible competitive overhang. Jensen quantified the impact: "we are gaining share tremendously fast in inference." The frontier-model partner roster is now complete; competitive losses to internal-ASIC architectures (AWS Trainium, Google TPU, Meta MTIA) are limited rather than expanding.

3. The New Segmentation Framework

Management restructured the disclosure into two market platforms (Data Center + Edge Computing) with Data Center split into Hyperscale + ACIE. The framework reflects management's view that NVDA's customer base has evolved beyond the "5-6 hyperscalers" framing into a structurally diverse mix of AI clouds, sovereign AI, enterprise on-prem, industrial, and physical AI / edge. Nine quarters of historical revenue posted in the new framework to enable Street modeling.

"The first thing, of course, is AI includes languages and depending on the different industries, it could be 3D graphics, manufacturing and industrial robotics… AI is diverse. The second thing is the applications are diverse… Where it runs is diverse… And then, of course, how it is governed… The easiest go to market, of course, is the hyperscaler because they are only you know, 5 or 6 of them. But the rest of the industry represents a couple of 250 thousand companies around the world."
— Jensen Huang, CEO

Assessment: The segmentation change is the single most consequential disclosure improvement for forward modeling. The Street had been blackbox-modeling NVDA as "hyperscaler revenue + everything else." The ACIE submarket disclosure ($37B / +31% QoQ) quantifies the diversification away from hyperscaler concentration — a structural concern that had been a bear talking point through FY26. The new framework also surfaces the sovereign AI contribution (+80% YoY) as the multi-year multi-country growth engine that's emerging behind the Anthropic + hyperscaler headlines.

4. Vera Rubin Production Shipments Commence Q3 FY27

Vera Rubin production shipments commence in Q3 FY27 with initial ramp; Q4 sees significant volume; Q1 FY28 expected to be larger still. The architecture: 7 purpose-built chips across 5 accelerated racks; 35x higher inference throughput vs. Blackwell; 10x greater AI factory revenue per dollar of capital deployed. Google's XGS bare-metal instances will support up to 960k Rubin GPUs across multiple sites.

"VeraRubin is going to be even more successful than Grace Blackwell at this point. Every single — I cannot think of 1. Every single frontier model company will jump on VeraRubin from the get go. And that was not true before on Blackwell. And so VeraRubin is off to a tremendous start and it will it will surely be more successful than even Grace Blackwell."
— Jensen Huang, CEO

Assessment: Vera Rubin is the next-generation architecture that should drive FY28 revenue growth. The 35x throughput / 10x revenue per AI factory framing is the key economic argument — at iso-CapEx, customers running Vera Rubin generate materially higher returns on their data center investment than Blackwell, which sustains and accelerates demand into FY28. Jensen's "supply constrained throughout the entire life of Vera Rubin" framing is the bull-case supply signal — demand is so high that NVDA cannot ship as much as customers want for 18-24 months. The Q3 FY27 production start, Q4 FY27 ramp, Q1 FY28 acceleration cadence is the most concrete forward visibility we have for FY28 revenue trajectory.

5. Agentic AI Has Arrived — The Demand Inflection

Jensen's closing remarks framed the call's central thesis: "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." The inflection from one-shot inference (chat) to reasoning (Q3 25 era) to agentic (current) is the structural demand catalyst — agents need orders of magnitude more compute than chat because they spin off sub-agents, use tools, do tool calls in parallel, and generate token-bursts during reasoning chains.

"In the AI era, compute capacity is revenue, and profits. NVIDIA is the platform of this era. Of all the platforms in the world, NVIDIA Compute supports the richest diversity of demand… The world is rebuilding computing for agentic AI and robotic physical AI. NVIDIA sits at the center of these transitions."
— Jensen Huang, CEO (closing remarks)

Assessment: The agentic-AI framing is the most important strategic thesis on the call. If correct, it changes the demand curve: each human user maps to 1 cloud query/day; each agent maps to thousands of queries/day; each agent will spawn sub-agents that each query thousands of times/day. Token consumption scales geometrically with agent adoption. Jensen's "billions of agents using PCs" framing puts the long-term compute TAM at multiples of current estimates. The Street's $3-4T annual AI infra spending by 2030 framing follows directly from this — and NVDA captures the majority of that spending if the platform thesis holds.

6. AI Factory Economics — $1T to $3-4T Annual Visibility

Multiple framings throughout the call quantified the demand backdrop: (a) hyperscaler CapEx forecast to exceed $1T by 2027 (analysts now forecasting); (b) AI infrastructure spending on track to reach $3-4T annually by end of decade; (c) $1T cumulative Blackwell + Rubin revenue visibility CY2025-2027 reiterated from GTC.

"With analysts now forecasting hyperscale CapEx to exceed $1 trillion by 2027 and Agentic AI beginning to proliferate all industries AI infrastructure spending is on track to reach $3 trillion to $4 trillion annually by the end of this decade."
— Colette Kress, CFO

Assessment: The $3-4T annual AI infra spending by 2030 framing is the most aggressive long-term TAM we have seen NVDA explicitly endorse. At current NVDA market share (~80-85% of AI accelerator + dominant networking + emerging CPU), the company captures the largest single share of that spending. If even 25-30% of the $3-4T flows to NVDA hardware (the rest to power, cooling, real estate, OEMs, ODMs), NVDA's late-decade revenue trajectory is $800B-1.2T annually — a structural multi-decade compounder framing the multiple hasn't yet incorporated.

7. AI Factory ROI — Rental Rates Rising, Asset Life Extending

The most underappreciated economic argument on the call: H100 cloud rental rates +20% YTD; A100 rental rates +15% YTD. Customers are "generating profitable revenue beyond the depreciable life of their GPUs." Combined with the 2.7x throughput increase + 60% token cost reduction on GB300 vs. 6 months ago (full-stack innovation), the AI factory economics are improving on multiple axes simultaneously.

"The vast and trusted marketplace for NVIDIA compute is a critical foundation on which billions in AI infrastructure spending is being financed by the ecosystem… The price of renting an H100 has risen 20% year to date, while A100 cloud pricing is up nearly 15%. Benefiting from the versatility of our platform and continuous performance enhancements enhanced by our software stack, customers are generating profitable revenue beyond the depreciable life of their GPUs."
— Colette Kress, CFO

Assessment: This is the most important counter to the "AI bubble" bear narrative. If H100 (a 4-generation-old product) is renting at +20% YTD and generating profitable revenue beyond depreciation, the unit economics of buying NVDA GPUs are materially better than the bear case assumes. This validates the $1T+ AI infrastructure spending pipeline because the ROI math actually works — customers are not over-paying for capacity; they're under-supplied for demand. Asset-life extension to 5-6 years (vs. typical 3-4 for compute infrastructure) directly compounds customer ROI.

8. China — H200 Approved But No Revenue in Outlook

US government has approved H200 licenses for China-based customers, but NVDA has not generated any revenue yet and is "uncertain whether any imports will be allowed into the country." Q2 outlook does NOT include any China Data Center compute revenue (consistent with prior quarter posture).

"While the US government has approved licenses for H200 to be shipped to China based customers, We have yet to generate any revenue. And we are uncertain whether any imports will be allowed into the country. As a result, consistent with last quarter, we are not including any China data center compute revenue in our outlook."
— Colette Kress, CFO

Assessment: China remains the binary risk and the binary upside. Current $91B Q2 guide includes ZERO China data center revenue. If China imports resume (license approval + customer demand intact), incremental upside is $5-10B/quarter at peak pre-restriction levels. The conservative posture of excluding it from guide preserves optionality — any China resumption is pure upside surprise. The "uncertainty around imports" framing suggests current US-China dynamics are not resolved; resolution probably 2H 2026 or later.

9. Capital Return Aggression — Dividend 25x, $80B Buyback

Q1 capital return $20B (record). Dividend raised from $0.01 to $0.25/quarter (25x increase — Jensen corrected Colette's $0.20 framing in real-time on the call). New $80B share repurchase authorization on top of $39B remaining = $119B total authorization. Plan to return ~50% of FCF to shareholders in FY27.

"With confidence in our long term free cash flow outlook, and our commitment to sharing our success with shareholders we are increasing our quarterly dividend from $0.01 to $0.20 per share."
— Colette Kress, CFO

"First of all, Colette meant to say we are increasing our quarterly dividends from $0.01 to $0.25. I think that extra $0.05 would mean a lot to the large shareholders."
— Jensen Huang, CEO (live correction)

Assessment: The Jensen live correction on the dividend amount is a small moment that captures something important — NVDA management is comfortable enough with the long-term FCF trajectory to take the higher number. The $0.25/quarter ($1.00 annual) at current share count is ~$24B/year of dividend (still tiny relative to $200B FCF run rate), but the 25x increase is the symbolic signal. Combined with the $80B incremental buyback authorization, capital return is now a material framework rather than an afterthought. At ~50% of FCF returned, NVDA is now in dividend-aristocrat-eligible cash-return cadence — the kind of capital allocation profile typically reserved for mature compounders, not 85%-YoY-growing semiconductor companies.

10. Sovereign AI — 40 Countries, $50T GDP

Sovereign revenue +80% YoY. NVDA AI infrastructure now deployed across nearly 40 countries representing $50T of GDP. Number of partner data centers exceeding 10MW has nearly doubled in one year to over 80 sites.

"Sovereign revenue increased more than 80% year over year. NVIDIA AI infrastructure is now deployed across nearly 40 countries representing $50 trillion in GDP."
— Colette Kress, CFO

Assessment: Sovereign AI is the under-appreciated multi-year growth engine. Each country building national AI infrastructure represents $1-10B of multi-year compute commitments, and the pipeline keeps expanding (40 countries today; trending higher). Sovereign AI revenue has lower competitive intensity (NVDA is the only full-stack solution provider for sovereign deployments at scale), higher margin (less price-sensitivity), and longer revenue tail (multi-year national programs vs. quarterly hyperscaler CapEx cycles). The 40-country footprint is itself a moat — building parallel sovereign capacity on TPU or Trainium would require multi-year ecosystem replication.

11. Supply Chain — $145B Inventory Purchase Commitments + Prepaids

Total supply (inclusive of inventory purchase commitments + prepaids) reached $145B in Q1. NVDA is "not immune to supply challenges" but "confident in our ability to support the growth opportunity ahead." Vera Rubin will be supply-constrained throughout its entire life.

"We remain front footed in securing sufficient supply to support our customers' growth. In Q1, we increased total supply inclusive of inventory purchase commitments on prepaids to $145 billion. While we are not immune to supply challenges, we remain confident in our ability to support the growth opportunity ahead."
— Colette Kress, CFO

Assessment: $145B of supply commitments is the strongest forward-revenue signal management has provided. At current ~$330B annual revenue and growing, the $145B supply commitment represents roughly 5-6 months of forward sell-through — meaning NVDA is structurally pre-committed for supply through CY2026 and partially into CY2027. The "supply constrained throughout the entire life of Vera Rubin" framing implies NVDA cannot exceed demand even with all of its current commitments deployed. This is the inverse of the typical semiconductor cycle bear case (excess capacity, inventory write-downs); NVDA is in a structural under-supply position with full commercial benefits.

12. Edge Computing + Physical AI

Edge Computing $6.4B (+10% QoQ / +29% YoY) covers PCs, gaming, workstations, AI RAN, robotics, automotive. Physical AI revenue exceeded $9B over the trailing 12 months. Uber robotaxi partnership will deploy across nearly 30 cities and 4 continents by 2028; leading companies across industrial, surgical, and humanoid robotics are building on NVDA technology.

Assessment: Physical AI is pre-monetization but pre-positioned. NVDA Jetson + Cosmos + Isaac + Drive platforms have captured the architectural standard for autonomous systems before the revenue inflection. The Uber robotaxi rollout timeline (30 cities / 4 continents / 2028) is the most concrete monetization milestone; humanoid robotics + industrial robotics are 2027-30 stories. Edge is not material to FY27 modeling but adds 10-15% of revenue by FY30 if physical AI inflects as Jensen describes. Optionality leg of the thesis.

Guidance & Outlook

The Q2 guide is the cleanest forward signal on the call. $91B midpoint above buy-side whisper $90B, with +11% sequential growth on what was already a record $82B base. Critically, the guide explicitly excludes China Data Center compute revenue — meaning even at zero China contribution, NVDA expects to add another $9B of net new quarterly revenue. The implied sequential growth comes "primarily from data center" with the new ACIE segment and continued hyperscale Blackwell deployment driving.

Implied FY27 setup: If Q2 lands at $91B and H2 sees Vera Rubin ramp + continued ACIE compounding + sovereign acceleration, FY27 revenue tracks $385-420B (vs. prior consensus ~$340B and prior Aardvark model ~$320B). This is a meaningful upward revision in a single quarter — Street estimates likely move from $340B FY27 to $375-395B within 1-2 weeks.

Street EPS: Pre-print FY27 consensus EPS was ~$6.50. Post-print, expect consensus to move toward $7.50-8.00 within 2 weeks on (a) Q1 + Q2 base running ahead, (b) Vera CPU $20B incremental revenue, (c) Anthropic partnership capacity buildout, (d) lower tax rate guide.

Guidance style: Disciplined-bullish. Q2 guide above whisper without overreaching; FY27 framework reiterated (mid-70s GM, $1T cumulative Blackwell+Rubin CY25-27, $3-4T annual AI infra by 2030). Management has earned the credibility to commit forward without hedging.

Analyst Q&A Highlights

The New Segmentation Framework — Hyperscale vs. ACIE

The dominant topic on the call. A recurring line of questioning probed the rationale for the new disclosure framework and the relative growth rates expected between Hyperscale and ACIE. Management's framing: the second segment (ACIE) is structurally larger and faster-growing than hyperscale despite a smaller current revenue base because (a) the customer count is hundreds of thousands vs. 5-6, (b) the addressable workload set is more diverse, (c) each customer requires NVDA's full-stack platform vs. the build-your-own approach hyperscalers can take.

Q: "Where do you put the neo clouds across those 2 segments? Are they in hyperscale, or are they in the AI cloud? Part of me assumes the latter, but I am not so sure. And then but just what is the magnitude of them? I mean, they are both about the same the same magnitude now. It almost sounded to me like you were suggesting that you thought the latter, the ACIE, cloud would grow faster. Maybe going forward than hyperscale."
— Stacy Rasgon, Bernstein Research

A: "You are correct. That AI native clouds do not build chips, do not design their own chips. And they do not wanna they cannot really assemble unrelated parts together into an AI factory. Their need for an architecture that has a great deal of offtake so that it runs every model, has customers from everywhere. Is incredibly high… If you look at that segment, it started growing after the AI ecosystem developed in the hyperscale… I expect the second category to be larger over time both in the near term over the next several years. I think it is a sure foregone conclusion. Both are gonna grow incredibly fast — I expect the second category to still grow faster."
— Jensen Huang, CEO

Assessment: ACIE structurally growing faster than Hyperscale is the most important strategic disclosure of the call. It diversifies NVDA's revenue mix away from hyperscaler concentration risk that has been a multi-quarter bear talking point. At +31% QoQ vs. Hyperscale +12%, ACIE will likely exceed Hyperscale revenue within 2-3 quarters if current trajectories hold. This re-rates the entire revenue mix story.

Growth Rate vs. Hyperscaler CapEx

A line of questioning probed whether NVDA can sustain growth rates above the underlying hyperscaler CapEx trajectory (forecast +90-100% CY26 per Street). Jensen's response was framework-level: "we should be growing faster than hyperscale CapEx" because ACIE + sovereign + edge + physical AI all grow faster than hyperscaler infrastructure spending.

Q: "Your data center business ex China grew about 120% in the quarter, and then you are guiding about 100. CapEx at the hyperscalers is forecast by many, including myself, to like, grow 90 to a 100% this year. And you talked about data center still on track to be $3 trillion to $4 trillion by the end of the decade. Was just wondering the goal for the company to grow faster than hyperscaler CapEx? Do are you still comfortable in kind of endorsing that view?"
— Ben Reitzes, Melius Research

A: "First of all, we should be growing faster than hyperscale CapEx. And the reason for that is illustrated by the segmentation that I just described… The first category is hyperscale. And the CapEx is at a trillion dollars. And it is growing. Towards the 3 to 4. The second category. The second category is all of the AI native clouds. They are regional, They are all over the place… The second category is extremely diverse. Instead of 6 or 7 companies representing the revenues associated with our first category, The second category is hundreds, thousands of companies, and in the future, it will be hundreds of thousands of companies."
— Jensen Huang, CEO

Assessment: Jensen's commitment to "growing faster than hyperscaler CapEx" is the most explicit multi-year growth framework he has provided. It's also defensible: ACIE is growing 2.5x faster than Hyperscale; sovereign AI +80% YoY; physical AI emerging. Even if hyperscaler CapEx growth moderates from 90-100% to 30-40% in CY27, NVDA's total revenue growth can stay above that pace on the diversification mix shift.

Vera CPU — TAM and Use Cases

An exchange on the Vera CPU disclosure produced the most quantitative forward commentary from management on the new revenue stream. Vera serves four use cases: (1) paired with Rubin in Vera-Rubin systems (priced separately), (2) standalone CPU, (3) Vera + CX9 + storage software stack, (4) Vera + CX9 + security/confidential computing. The $20B FY27 visibility is standalone-only.

Q: "There is a lot of excitement around CPU for Agentic applications and just a lot of noise around the number of CPUs actually exceeding the number of GPUs. And just hoping that you could kind of give your perspective that, first of all, you know, is this an incremental workload? Is this kind of cannibalizing what the GPU would have done otherwise? And then secondly, the $20 billion number that you gave, is that for stand alone Vera CPUs?"
— Vivek Arya, Bank of America

A: "The 20 billion is for standalone CPU. And remember, we have Vera is used in 4 ways… The world has a billion human users, My sense is that the world is gonna have billions of agents. Not today. I mean, we are gonna grow into it. But we will have billions of agents… The economics of AI of the future is tokens per dollar. Or dollars per token. And so what we need to do in the future is to generate tokens process tokens as fast as possible. And that is what Vera does incredibly well."
— Jensen Huang, CEO

Assessment: Vera CPU $20B FY27 standalone revenue is the most quantitative new datapoint from the call. The "billions of agents" framing puts the long-term CPU TAM at multiples of the current $200B disclosure — every agent has a CPU footprint for orchestration, tool use, memory management. At current x86 CPU market share (Intel + AMD ~95%), NVDA capturing even 5-10% of the agentic CPU socket is $10-20B incremental annual revenue.

Inference Share Gains — Vera Rubin Positioning

A question on inference market share dynamics elicited the strongest competitive positioning statement of the call. Jensen was explicit that "we are growing share in inference, and we are growing share in inference very, very quickly" — driven by the addition of Anthropic to the partner roster, the launch of new frontier-model companies (Cursor, Perplexity, TML, Reflection, etc.), and Vera Rubin being adopted from day-one by frontier-model providers.

Q: "Vera Rubin coming soon, and you obviously have great insight into coming updates to Frontier models, new techniques to optimize around diverse AI workloads with investors keenly focused on your market share and inference, how do you see Vera Rubin in your extreme co engineering impacting your share of the inference market, you know, as we look into late 26, 27?"
— Chris Muse, Cantor Fitzgerald

A: "We are growing share in inference. And we are growing share in inference very, very quickly… We added Anthropic to our partnership this year. They are expanding incredibly fast. We have partnered with them to secure computing capacity across Azure, AWS, CoreWeave, I forget who else we have already announced, but there is a whole list of others that we are bringing online for them… Vera Rubin is going to be even more successful than Grace Blackwell at this point. Every single — I cannot think of 1. Every single frontier model company will jump on Vera Rubin from the get go. And that was not true before on Blackwell."
— Jensen Huang, CEO

Assessment: "Every single frontier model company will jump on Vera Rubin from the get-go" is the most aggressive positioning statement Jensen has made in any earnings call. The implication: ASIC alternatives (AWS Trainium, Google TPU, Meta MTIA) are not displacing NVDA at the frontier-model layer; they're filling specific cost-optimization niches within hyperscaler workloads. The frontier-model layer — where the highest-margin compute spending happens — is fully NVDA.

LPX / Specialized Inference Accelerators

A line of questioning probed traction with NVDA's specialized inference products (LPX, CPX) vs. the general-purpose GPU + CPU stack. Jensen's framing: LPX is designed for low-latency, high-token-rate, premium-token services with low context-processing requirements — a niche product for some time. Mainstream inference + training + agentic workloads remain on the full GB / Vera Rubin platform.

Q: "I wanted to ask about the traction you are getting with some of these custom merchant things you are doing, stuff like, you know, CPX and LPX. And I just wanted to ask and see sort of you have talked before about fat synthesis being, I think, 20% of the market. So would imagine you are getting pretty good traction with LPX. So can you just talk about that and maybe also how that fits into your broader platform strategy?"
— Tim Arcuri, UBS

A: "The LPX is designed for low latency and high token rate. Its throughput is low, Size capacity is low. And it is context processing, its ability to absorb a lot of context, for example, for software coding, for agentic workloads, its ability to absorb a great deal of context is lower… I expect that LPX and other SRAM based decode focus, token high token rate generate focused accelerators will be a niche product for some time. For some time to come. As you know, Grace Blackwell and Vera Rubin we support the entire life cycle of AI from the data processing, preparing for training, to pretraining, to post training reinforcement learning all the way to inference."
— Jensen Huang, CEO

Assessment: Jensen's positioning on LPX as a niche product validates NVDA's full-stack GPU + CPU platform approach. The bear case has been that purpose-built inference accelerators (LPX-like ASICs from Groq, Cerebras, SambaNova) take share at the inference layer; Jensen's framing limits that share to specific premium-token use cases (10-20% at most of inference) rather than the broader inference market. The full-stack platform remains the dominant approach.

Anthropic Partnership Scale

A line of questioning on Anthropic capacity additions revealed the magnitude of the partnership shift. NVDA's coverage of Anthropic was "largely zero until just recently" — and capacity additions for Anthropic this year and next "will be quite significant. Very significant."

Q: "Could you maybe give us a sense about whether you expect that your CapEx for next year will be toward the lower end of the range you sort of outlined at the beginning of this year? Just any update on just how to think about [growth] in light of the slower recovery you seem to be talking about right now."
— Multiple analysts incl. Joseph Moore, Morgan Stanley

A: "We added Anthropic to our partnership this year. They are expanding incredibly fast. We have partnered with them to secure computing capacity across Azure, AWS, CoreWeave, I forget who else we have already announced, but there is a whole list of others that we are bringing online for them. So the amount of capacity that we are gonna bring online for Anthropic this year and next year is going to be quite significant. Very significant. And so we are growing and our coverage of anthropic has been largely zero until just recently."
— Jensen Huang, CEO

Assessment: Anthropic going from "largely zero" NVDA coverage to "quite significant. Very significant." capacity additions is a single-quarter share shift comparable in magnitude to capturing a hyperscaler workload. Anthropic Q1 26 revenue run rate was reported at $5B+ growing; if Anthropic's compute spend reaches $10-15B annual by FY28 and NVDA captures the majority, that's $5-10B of incremental Data Center revenue NVDA was not modeled for. The Anthropic shift is one of the largest single-customer share gains in recent NVDA history.

What They're NOT Saying

1. No quantification of FY28 framework: Despite the Vera Rubin Q3 production start commentary, no specific FY28 revenue range provided. Management is keeping FY28 framework for GTC and capital management forum.
2. No China revenue framework even in upside scenario: Even with H200 licenses approved, NVDA did not provide an "if China resumes" sensitivity. Conservative posture preserves optionality.
3. No specific Vera CPU customer disclosure: $20B FY27 standalone visibility and "every major hyperscaler and system maker partnering" but no named customer logos for Vera CPU specifically.
4. No quantification of the Anthropic capacity buildout dollar size: Multiple "quite significant. Very significant." framings but no $-magnitude. Could be $5B or $20B of incremental FY27-28 revenue.
5. No commentary on internal-ASIC competition (Trainium / TPU / MTIA): Despite Anthropic-from-zero framing, no direct address of how hyperscaler internal-silicon programs are progressing. Likely deliberate — surface only the positive share gains.
6. No specific Rubin ASP commentary: 35x throughput / 10x AI factory revenue framing is provided but no Rubin system ASP or gross margin guide. Implies Rubin ramp economics are similar-or-better than Blackwell but no commitment.
7. No update on power constraints at hyperscaler customer sites: Despite "power and capital constrained" framing in prepared remarks, no specifics on which customers are power-limited vs. supply-limited.
8. No commentary on the DSO bounce-back to mid-50s in Q2: The Q1 DSO of 45 days reflected favorable timing of collections; Q2 expects mid-50s. The 10-day swing implies a meaningful working capital build in Q2 that will reduce FCF conversion vs. Q1.

Market Reaction

• Pre-print setup: Stock close May 20 ~$220-225. YTD CY26 +5%; trailing 12M +45%. S&P rallied 600 points pre-print on positioning. Buy-side whisper: $80B revenue / $90B Q2 guide. Options-implied move ±7%.
• After-hours: Initial flat-to-slightly-down despite blowout numbers. Settled near $224, roughly flat. "Sell the news" dynamic on extremely high expectations.
• Sector read-across: MU +2% on HBM demand reaffirmed. ASML, AMAT, LRCX, KLAC +0.5-1% — tooling demand intact but no incremental upside. CoreWeave +3% on Anthropic capacity buildout. ANET +1.5% on InfiniBand/Spectrum-X +4x YoY. Hyperscalers MSFT/AMZN/GOOG flat-to-up.
• Volume: Elevated but not extreme — sentiment positioning had been long-leaning into print.

The roughly flat reaction (~+0%) is the canonical sell-the-news dynamic in a name where buy-side expectations had moved well ahead of sell-side consensus. Street consensus was $78.75B revenue; actual was $81.6B (a 3.6% beat). But the buy-side whisper was $80B, so the actual print only modestly cleared that bar. Similarly, the Q2 guide $91B cleared sell-side consensus by 5% but only cleared the whisper ($90B) by 1%. When numbers clear sell-side by 3-5% but whisper by 1-2%, the price reaction tends to be muted — there's no surprise factor for incremental buyers, and existing longs have less reason to add.

The deeper signal in the muted reaction is what the market is NOT yet pricing: the Vera CPU $200B TAM disclosure ($20B FY27 standalone visibility entirely new), the Anthropic partnership at scale ("very significant" multi-year capacity buildout from "largely zero"), the new ACIE segmentation showing the diversification story has more legs than hyperscaler-concentration models assumed, and the FCF compounding ($49B in Q1 alone, ~$200B annualized) supporting the $80B incremental buyback authorization. All four are structural positives that get re-priced over weeks rather than after-hours, and we expect FY27 EPS consensus to move from ~$6.50 to $7.50-8.00 within 2 weeks as sell-side digests Vera CPU + Anthropic + ACIE.

The bear narrative this print needed to crack was "demand peaks H2 FY27 / Blackwell digestion." The combination of (a) Q2 guide $91B above-seasonal, (b) "supply constrained throughout entire life of Vera Rubin" framing, (c) Anthropic capacity additions for "this year and next year," and (d) sovereign AI +80% YoY across 40 countries collectively dismantles that thesis. Digestion isn't coming because the demand surface keeps expanding faster than NVDA can supply. The new bear narrative would have to be "rate of growth deceleration into FY28" rather than "absolute demand peak" — a meaningfully less compelling bearish framing.

Street Perspective

Debate: Is the Stock Already Priced for Perfection?

Bull view: The bull case argues that at $224 / ~29x FY27 EPS ($7.75 mid-estimate), NVDA is trading at a mid-quartile multiple on rapidly upward-revising estimates. With FY27 EPS still climbing toward $8 on Vera CPU + Anthropic + ACIE acceleration, and FY28 EPS power now plausibly $11-13 on Vera Rubin ramp + Vera CPU scaling + sovereign continued, the forward multiple is 17-20x on FY28 — historically the bottom-decile for NVDA in a growth-acceleration regime. Fair value supports $280-320 within 12 months as estimates catch up.

Bear view: The bear case argues NVDA at $224 is already pricing perfection. Any deceleration from the +85% YoY pace would compress the multiple meaningfully; ASIC competition (Trainium, TPU, MTIA) eventually displaces some hyperscaler workloads; the "billions of agents" narrative is asymmetric to disappointment; China remains a structural overhang. The stock has rallied 45% over 12 months; the easy money is made.

Our take: The "priced for perfection" framing requires assuming current estimates are correct; in fact, current estimates are running materially below the actual trajectory. Vera CPU + Anthropic + ACIE + sovereign + Vera Rubin Q3 ramp are all elements not yet in Street FY27 numbers. The stock isn't priced for perfection; it's priced for the prior consensus. As consensus moves higher over coming weeks, the stock follows. The multi-quarter setup remains favorable.

Debate: Does ACIE Justify Multiple Expansion?

Bull view: The bull case is that ACIE structurally diversifies NVDA's revenue mix away from hyperscaler concentration — historically a multiple-compression overhang. At $37B / +31% QoQ and growing 2.5x faster than Hyperscale, ACIE will likely exceed Hyperscale revenue within 2-3 quarters. The customer base diversification (hundreds of thousands vs. 5-6) is structurally durable and warrants higher quality-multiple on the revenue stream. This is the SAP-vs.-AWS framing: ACIE is the "enterprise software" multiple model; Hyperscale is the "cloud infrastructure" multiple model.

Bear view: The bear case is that ACIE's diversification benefit is overstated because the underlying demand driver is still hyperscaler-adjacent (AI clouds are essentially hyperscaler tier-2). And the +31% QoQ growth rate is unsustainable at scale — ACIE will moderate as the AI cloud build-out matures. The hyperscaler-concentration concern is real but cyclical, not structural.

Our take: ACIE diversification is real and structurally durable. The customer count (hundreds → hundreds of thousands) is the moat — AI clouds, sovereign AI, on-prem enterprise, industrial AI infrastructure are categorically different from hyperscaler concentration. The multiple-expansion case is justified by the mix shift, not just the absolute growth rate. The market hasn't yet processed this; it's a multi-quarter re-rating catalyst.

Debate: Vera CPU vs. AMD / Intel

Bull view: The bull case is that Vera CPU is purpose-built for agentic AI workloads (token throughput optimization) in a way that x86 CPUs (general-purpose, cores-per-dollar optimization) cannot match. The 1.5x performance per core / 2x performance per watt / 4x density specifications are not incremental improvements; they're structural advantages enabled by ground-up codesign with Rubin GPUs and NVLink. NVDA captures 10-20% of the $200B CPU TAM within 3-5 years — $20-40B incremental annual revenue.

Bear view: The bear case is that AMD (EPYC) and Intel (Xeon) have dominant share + deep customer relationships + mature software ecosystems. Vera CPU at FY27 launch is a 1.0 product competing against Gen 5 EPYC and Granite Rapids Xeon — execution and ecosystem are both meaningful disadvantages. AMD Bergamo + Turin EPYC already optimize for high-density cloud workloads; Intel Sierra Forest does similar. NVDA's $20B FY27 visibility is achievable but $200B TAM capture is years out.

Our take: Vera CPU is the most asymmetric optionality in the NVDA thesis. The downside is contained ($20B visibility well-modeled by NVDA); the upside is structural ($50-80B annual run rate by FY28 if agentic AI inflects). Even at the bear case, Vera CPU adds 5-7% to FY27 revenue. At the bull case, it's a meaningful new revenue engine that re-rates the multiple. Risk-adjusted return is highly positive.

Model Update Needed

Valuation impact: Fair value range raised to $250-310 (from prior $145-180 at Q1 FY26 plus interim updates). At $224 post-print, the upside is +12-38% over 12 months on continued earnings revisions and gradual multiple expansion as Vera CPU + ACIE thesis crystallizes. Catalysts: GTC Taipei keynote June 1, 2026; Computex; TD Cowen TMT conference May 28; BofA Global Tech June 4; Q2 FY27 print August 26, 2026.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened. Eight bull points confirmed (including three new); two bears reversed; three open. The single most important development is the structural diversification disclosure (ACIE +31% QoQ) plus the Vera CPU $200B TAM, which together reframe NVDA from "GPU monopoly with hyperscaler concentration" to "diversified accelerated-compute platform with multiple growth engines." This is the kind of strategic re-framing that takes 2-4 quarters for the multiple to fully price.

Action: Maintaining Outperform. Fair value range raised to $250-310. The +0% reaction is positioning, not thesis. Recommend holding existing positions through the GTC / Computex / TD Cowen / BofA conference circuit (June 1-4) for incremental positive catalysts; Q2 FY27 print August 26 becomes the next major test of the Vera Rubin Q3 ramp + Vera CPU early customer wins narrative.